I am trying to construct a time series of NDVI derived from Landsat 5, 7 and 8 using the Google Earth Engine's Python API. To ensure the validity of this TS, I am harmonising the TM/ETM+ imagery to OLI using the coefficients found by Roy et al. (2016, https://doi.org/10.1016/j.rse.2015.12.024), following the method outlined on https://developers.google.com/earth-engine/tutorials/community/landsat-etm-to-oli-harmonization. As the NDVI TS is pretty central to my research question, I would like to quantitatively assess the performance of this harmonisation using the OLS regression approach used by Roy et al. However, this requires overlapping pixels from different paths at similar dates to be extracted.

Is there a way of extracting this in GEE?

  • 1
    I have done this recently; started in the GEE but ended up downloading a set of images for my study region and doing it outside of the cloud. Basically you need to map a function over an image collection that will regress pixel values against images from the adjascent WRS rows. You will have to either filter an image collection to find the "companion image", or to create an image collection from a list of curated images (using .fromImages) - the latter being the brute force appraoch. Good luck.
    – korndog
    Jan 16, 2021 at 5:04
  • If you haven't already, you might also want to consider other ways to evaluate the success of the harmonization, e.g. using pseudo-invariant features or MODIS, before re-doing the analysis from the Roy paper.
    – korndog
    Jan 16, 2021 at 5:15
  • Hi cheers, for the advice. I'll have a look into those other methods I wasn't aware of them
    – joe_agate1
    Jan 16, 2021 at 9:42


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.